From 99531cf454bf63195197b0a2e839d184f4b160b0 Mon Sep 17 00:00:00 2001 From: Ronald Tse Date: Wed, 30 Sep 2026 22:04:41 +0800 Subject: [PATCH] release: ara-diac-small-int8static-2.1 (int8, parity cer_delta 0.1038pp on 2480 samples) --- models.yaml | 13 +++++++ ...ra-diac-small-int8static-2.1.metadata.yaml | 39 +++++++++++++++++++ 2 files changed, 52 insertions(+) create mode 100644 models/ara-diac-small-int8static/ara-diac-small-int8static-2.1.metadata.yaml diff --git a/models.yaml b/models.yaml index 667cb80..a7f1375 100644 --- a/models.yaml +++ b/models.yaml @@ -369,3 +369,16 @@ models: - {name: der_student_fullset, value: 4.5701, source: interscript/interscript-ml docs/RESULTS.md#run-006-r7-muon} parity: {samples: 2480, cer_delta: 0.0714} license: BSD-3-Clause + ara-diac-small-int8static-2.1: + task: diacritization + scripts: [Arab, Arab] + precision: int8 + filename: ara-diac-small-int8static-2.1-int8.zip + url: https://github.com/interscript/interscript-ml/releases/download/ara-diac-small-int8static-2.1/ara-diac-small-int8static-2.1-int8.zip + sha256: 7836bfd5582ad054ee419bfbd45a2b4a87fe8dc91e1f849898428faefd0c95da + size: 515362212 + metrics: + - {name: der_teacher_fullset, value: 2.289, source: interscript/interscript-ml docs/RESULTS.md#run-006-r7-muon} + - {name: der_student_fullset, value: 4.5701, source: interscript/interscript-ml docs/RESULTS.md#run-006-r7-muon} + parity: {samples: 2480, cer_delta: 0.1038} + license: BSD-3-Clause diff --git a/models/ara-diac-small-int8static/ara-diac-small-int8static-2.1.metadata.yaml b/models/ara-diac-small-int8static/ara-diac-small-int8static-2.1.metadata.yaml new file mode 100644 index 0000000..4069763 --- /dev/null +++ b/models/ara-diac-small-int8static/ara-diac-small-int8static-2.1.metadata.yaml @@ -0,0 +1,39 @@ +format: imf-v1 +id: ara-diac-small-int8static-2.1 +task: diacritization +source_script: Arab +target: Arab +tokenizer: bytes +opset: 14 +decoder: kv +precision: int8 +license: BSD-3-Clause +trained_from: 'sequence-level KD from the r7 canonical teacher (rababa_arabic_byt5/run-007-news/best, + 2.2864 windowed DER-CE full protocol): fresh greedy r7 labels on the same r5-units + corpus/limits as ara-diac-small-1.0, Muon optimizer (E3-adopted), vanilla ByT5-small + (E4, pre-registered gate <= 6.26). Checkpoint rababa-checkpoints:/rababa_arabic_distill_small/run-007-r7-muon-6ep/best. + The two measured wins compound: 8.259 -> 4.8218 full-set windowed DER-CE (teacher + reproduces 2.289 in-run vs documented 2.2864) — a 42% error reduction on the 1.0 + release at the same architecture and artifact size. Still misses the strict teacher+0.5pp + gate (+2.53pp; miss disclosed); the E2/E3 factorial attributes the residual to domain + coverage.' +metrics: +- name: der_teacher_fullset + value: 2.289 + protocol: windowed DER-CE (1400-byte windows, word-boundary split, greedy, haraqat-projected, + Misraj evaluator); full 1,200-paragraph SadeedDiac-25; in-run reproduction of + the documented 2.2864 (r7 canonical teacher) + source: interscript/interscript-ml docs/RESULTS.md#run-006-r7-muon +- name: der_student_fullset + value: 4.5701 + protocol: same full-set harness; E4 (r7 teacher labels + Muon, vanilla ByT5-small) + vs the 8.259 AdamW/r6-labels 1.0 release — a 42% error reduction at identical + architecture and artifact size + source: interscript/interscript-ml docs/RESULTS.md#run-006-r7-muon +parity: + samples: 2480 + cer_delta: 0.1038 +sha256: + decoder-kv.onnx: f56a365df3c7acefcc96982ab3fb70d7c417aea9a4379bd1a39a784dc227db98 + decoder.onnx: 81852f41838f6ad8dd8ee5b72e709328afdad474ee28da6381abbd32659219d7 + encoder.onnx: 61afa9ded2e61966a2966f906650715183f95cccc24bac699ed4c4b2d766eecd